Skip to main content
AI & Agents · Programming · Algorithms & ML · Data · Ages 9-14 · No experience needed

Build your own AI.Code it like a game.Level up your model.Win with data.

Deep play, real learning: every session ends with something that actually runs.

From their first line of code to their first AI-powered creation, children learn through projects — building, testing, questioning and explaining as they go.

Disciplines
4
Age
9-14
Rote learning
0%
Builds kept
100%
01 / Subjects

What we teach

Every subject is taught as a project: students make something, then explain the thinking behind it.

  • SUBJECT 01 · Prompt · Connect · JudgeLive

    AI & Agents

    Understand how a language model produces an answer, learn to write clear prompts, then connect AI to tools to build an agent that completes a task on its own, and decide which calls should never be left to it.

    • Language models
    • Prompt design
    • Tool use
    • AI safety & ethics
    1. 01 · Prompt with intent

      State the goal, limits and format so the AI does what was asked.

    2. 02 · Build an agent

      Let the AI call tools and finish a real task step by step.

    3. 03 · Check the AI

      Verify its output and tell which decisions must stay with people.

    DeliverableAn AI helper that completes a task with a human check built in
  • SUBJECT 02 · Decompose · Code · DebugComing

    Programming

    From blocks to Python: children break a task into steps, write programs that actually run, and learn to find and fix their own bugs.

    • Block coding
    • Python
    • Minecraft
    • micro:bit
    1. 01 · Break it down

      Turn a big task into small steps a computer can follow.

    2. 02 · Code it

      Write code that actually runs, not a copied example.

    3. 03 · Debug it

      When it breaks, find which step went wrong before changing anything.

    DeliverableA working program they wrote and debugged themselves
  • SUBJECT 03 · Compare · Train · EvaluateComing

    Algorithms & Machine Learning

    From searching and sorting to training a first model: children compare how fast different solutions are, then collect data, train a model, and see what it learned and what it got wrong.

    • Search & sort
    • Pathfinding
    • Classifiers
    • Evaluation
    1. 01 · Compare solutions

      Solve one problem two ways and count which is faster.

    2. 02 · Train a model

      Collect and label the data and train it, rather than calling a ready-made result.

    3. 03 · Find why it fails

      When the model is wrong, go back to the data and find what taught it the mistake.

    DeliverableA model they trained and can explain
  • SUBJECT 04 · Collect · Clean · VisualiseComing

    Data Processing

    Collect real data from sensors and spreadsheets, clean and organise it, chart it, and use it to answer a question they asked themselves.

    • Sensor logging
    • Spreadsheets
    • Data cleaning
    • Charts
    1. 01 · Ask a question

      Decide what to answer before deciding what to collect.

    2. 02 · Clean the data

      Find missing, duplicate and obviously wrong records before analysing.

    3. 03 · Show it in a chart

      Pick the right chart and make the finding clear to someone else.

    DeliverableA chart-based report that answers a question with their own data
AGES 9–12 · NO EXPERIENCE · 12-LESSON PROJECT

AI Guardians: Rebuild the Lost City

From a smart house that opens only for its family to a code-built bridge and a digital twin in Minecraft, and a child who can explain where the AI falls short.

  1. 01
    BUILD

    Build a house, then a city

    Make a wooden smart house greet visitors and obey gestures, then use loops, functions, and variables to build walls, a bridge, and a coaster in Minecraft.

  2. 02
    TRAIN

    Train a model by hand

    Train a gesture remote on a micro:bit and a camera door lock that knows the family.

  3. 03
    TEST

    Find the AI's blind spots

    Record the lock's mistakes, check the careless AI Architect's blueprints, and discover how an agent can exploit a reward rule.

  4. 04
    EXPLAIN

    Take responsibility for results

    Explain how the system works, where it may fail, and which decisions must stay human.

Course specai-guardians
Age
Ages 9-12
Length
12 lessons · 90 min per lesson
Language
Taught in Chinese. English student decks. Bilingual worksheets.
Format
In person / Fully online
Prerequisites
No coding, Minecraft, or AI experience required

> AI & Agents · students keep what they build

02 / Pedagogy

How we teach

  1. Make it first, explain it second

    Children make something that actually works first, then go back and understand why it holds. Explanation serves the task in their hands, not the other way round.

  2. Every session ends with something to show

    Each session ends with something a child can show someone. Parents do not have to wait until the end of term to see what was learned.

  3. You have learned it when you can explain it

    Getting it right is not the same as understanding it. Every session asks a child to say it in their own words, and where the words fail is where the learning is missing.

  4. A mistake is not finished until it is diagnosed

    We do not just hand over the right answer. A child has to point at which step went wrong and why they thought that, or the same mistake comes back.

03 / Fit

Who thrives here, and who might not

Thrives here

  • Complete beginnersNo coding, hardware or AI experience assumed. Courses build the foundation from the first session rather than assuming prior practice at home.
  • Children already coding a littleEvery task carries a stretch challenge, so a student with some experience is neither idle nor held back waiting for the rest of the class.
  • Children who want a future in AIChildren train models, build agents and check where an AI goes wrong: an early taste of how people who work in AI actually work, and of which decisions must stay with humans.

Might not

  • Looking for an easy way to fill an afternoonSessions ask for sustained effort and self-directed problem solving. If a family wants relaxed supervision, a child will find this hard going.
  • Needs marks and rankings right awayWe assess what a child can build and explain, and we do not rank. Families who want weekly score comparisons will find the feedback too slow.
  • Would rather not build anythingEvery session involves making something, getting it wrong, and doing it again. A child unwilling to build will not enjoy it.
04 / For parents

What parents ask: direct answers

Do parents need to buy expensive kits or 3D printers?

Absolutely not. All the tools, microcontrollers, components and measuring instruments used in class are provided. When a term project is finished, students take their completed, working build home to keep, with no surprise kit fees.

Policy: all studio hardware included

How are child photos and learning records handled?

We follow Malaysia's PDPA. Photos of student work and recordings of their presentations are shared only with parents, and we never use a child's face in public marketing without consent.

Policy: zero unconsented commercial use

What happens when a child breaks an expensive part?

We celebrate the learning moment. Things break when you explore limits. Parents are never charged a replacement penalty; our teachers use the broken part or fried circuit as a specimen for finding the root cause.

Policy: no replacement surcharges

05 / FAQ

Frequently asked

How many students in a class?

Small classes of 8 to 12 students. The teacher can follow every child’s progress and questions, and children find it easier to discuss ideas and build projects together.

What if my child misses a session?

Please let us know at least 24 hours ahead. We will first try to place your child in another class at the same point in the course, subject to space. If no suitable class is available before the next scheduled lesson, we send you the lesson materials and task, and the teacher helps your child catch up on the key steps next time. The number of make-up lessons per term is set out when you enrol.

What does my child have at the end?

Students keep their own code, project files and physical builds, and receive a competency-based record of what they can actually do rather than a certificate of attendance.